container-debug
🩺 Container debug
Most container failures are not mysterious. The runtime records why a container stopped — the exit code, whether the kernel's OOM killer fired, how many times it restarted, what its healthcheck printed — and most of the rest are a short list of failures that only exist because the process is in a container: localhost meaning the container itself, a binary built for the wrong architecture, an entrypoint with Windows line endings, a non-root user that can't write where the app expects, a signal that never reaches the app. This skill reads that record first, matches it against that list, and proves the cause before anyone changes a line.
Where this fits
The container skills; this one is called when something built with the others misbehaves:
- dockerfile — the image
- docker-compose — the local stack
- container-hardening — how containers run, securely
- container-debug (this skill) — why one isn't running the way it should
- docker-swarm — the cluster and stack files; on Swarm,
docker service ps --no-truncis where this skill's Step 1 starts
Where the fix lands decides who makes it: an image defect goes to dockerfile, wiring between services to docker-compose (or the platform's manifests), a security setting to container-hardening, and a bug in the application itself to hotfix. A production outage is an incident first — hand it to incident-fix, which contains before it diagnoses, and use this skill's tables as its evidence-gathering.
Reference docs: